nesting and convective systems · project statement: “the current nesting techniques in hwrf and...
TRANSCRIPT
Slide 1 of 29
Vijay Tallapragada
NOAA/NWS/NCEP/EMC
Nesting and Convective Systems Update on Team Plans and Activities
Next Generation Global Prediction System
(NGGPS)
HIWPP/NGGPS Program Status Meeting
February 9-10, 2016
Slide 2 of 29
• Chair: Vijay Tallapragada, EMC
• Members: – EMC: Tom Black, Samuel Trahan, Dusan Jovic, Matt Pyle, John
Michalakes, Bin Liu
– AOML: S.G. Gopalakrishnan, Thiago Quirino, Steven Diaz
– GFDL: S.J. Lin, Lucas Harris, Morris Bender, Tim Marchok
– ESRL: Stan Benjamin, Jin Lee, Ligia Bernardet
– NCAR: Bill Skamarock, Chris Davis
– Navy: Jim Doyle
– PSU: David Stensrud, Paul Markowski, Yvette Richardson
– U. Michigan: Christiane Jablonowski, C.M. Zarzycki
NGGPS Nesting/Convective Systems
Team Membership
Slide 3 of 29
• Incorporate more sophisticated nesting or mesh
refinement capabilities in the NEMS framework
• Development of generalized nesting or mesh refinement
techniques
• Implement multiple static and moving nests globally, with
one- and two-way interaction and coupled to other
(ocean, wave, sea ice, land, etc.) models using NEMS
infrastructure
• Implement scale-aware physics appropriate for the high-
resolution nests
• Post-processing, product development and verification of
high-resolution model output
NGGPS Nesting/Convective Systems
Team Objectives
Slide 4 of 29
• Strategic development approach
– Are the generalized nesting techniques truly independent
of choice of NGGPS dycore(s)?
– How scalable and efficient will be the global models with
two-way interactive nests for operational considerations?
(interactions with overarching system/software
architecture and engineering teams)
– Need for developing appropriate physics and initialization
techniques (interactions with atmospheric physics and
data assimilation teams)
– Need for developing advanced diagnostic and verification
tools for evaluating truly non-hydrostatic model forecasts
at cloud resolving scales
NGGPS Nesting/Convective Systems
Team Approach
Slide 5 of 29
– Take advantage of already developed (and ongoing
developmental) work in the HWRF and NMMB/NEMS
systems
– Accelerate design and development of efficient two-way
interactive nests using generalized nesting framework
using ESMF/NUOPC coupler functionality in NEMS
– Design of Adaptive Nests that can be activated,
terminated, (or re-activated) during model integration
Near term Challenges
Slide 6 of 29
• Static/moving
• 1-way/2-way interactive (nests)
• Multiple nests run simultaneously
• Bit reproducible and restartable (static/moving/
1-way/2-way )
• Very fast and efficient!
• Dynamics, physics and initialization appropriate
and applicable for high-resolution nests within
the global model
General Requirements for Operational
Nesting or Grid Enhancement
Slide 7 of 29
Generalized Nesting By Coupling
NGGNF can be considered as a subset of the “Coupled NEMS” project. While “Coupled NEMS” focuses on coupling various earth system models through surface interactions, NGGNF is a dynamic interface for nesting that couples one or more atmospheric models at the dynamic level by providing several required numerical transformations across the interface, including 3D interpolations.
NGGNF addresses the shortcomings of the current generation of nesting technology:
1. Offers a high degree of portability and abstraction
• A complete, stand-alone nesting algorithm
• Built into NEMS framework
• No dependence on dynamical core, grid shape, grid projection, or vertical coordinate
2. Offers advanced inter-domain interaction features
• Nests can interact with scale spanning resolution of the parent domain
• Nests can move across domain edges and poles without distortions
• Supports sequential/parallel, 1 and 2-way interactive, domain integration
3. Offers novel features
• Generalized IO: eliminates custom model IO code and performs post-processing
AOML in partnership with EMC and other OAR labs is building the Next Generation Generalized Nesting Framework (NGGNF) within NEMS to advance global-2-local scale modeling for hurricanes
Architecture of the NEMS NUOPC “mediator” with the NGGNF dynamic layer
NGGNF
The Coupled NEMS Project
NGGNF
Slide 8 of 29
Project Statement: “The current nesting techniques in HWRF and NMMB are based on the projection center of the parent grid, limiting their applications
to a confined region in the tropics, and limiting their ability to scale well at higher resolutions and pole-ward locations. A generalized nesting, core
independent nesting technique that can work independent of the parent model’s grid structure as well as map projections will advance the state-of-the-
art in nesting techniques (one-way as well as two-way).
SVN Repository path: https://svnemc.ncep.noaa.gov/projects/hnmmb/branches/AOML-HRD/NGGNF
Current Framework (2016)
NEMS Framework
Example of Core, Grid, and Projection Independent, dynamical (up/)downscaling using and advancing ESMF re-gridding
Model 2 Model 3 Model 5
Model 1
Model 4
Model 3
Model 2
Model 4
Model 5
Example: Generalized Nesting By Coupling
Slide 9 of 29
• Developed a Hurricane NMMB repository at EMC
– Transitioned HIWPP funded Idealized TC framework to repository
https://svnemc.ncep.noaa.gov/projects/hnmmb
– Currently transitioning Basin-scale HWRF multi-storm initialization into NMMB
Configuration
1) Resolution: 18:06:02km,
61 levels, 2mb model top.
2) Initial TC intensity: 20 m/s
at 1002 mb.
3) Physics package: HWRF
with high-frequency calls
Idealized TC Framework in NMMB
Hurricane Developments in NMMB (EMC-HRD Collaborations supported by HIWPP and R2O/NGGPS)
Slide courtesy: Thiago Quirino, HRD/AOML
Slide 10 of 29
Efficiency and Scalability of NMMB (EMC-HRD Collaborations supported by HIWPP and R2O/NGGPS)
• Evaluating NMMB scalability with 2-way nest
interaction
– Performed various timing experiments: 1-way
and 2-way interaction, varying grid sizes,
multiple nests, and HWRF physics package
– Determined that scalability limitations are
similar to those of HWRF:
• Scalability efficiency levels-off as tile size
reaches ~ 12x12 points
• Halo exchanges and collective MPI calls in
solver are costly
• Frequent physics calls (for high resolution
forecasts) are costly
• Forcing and feedback costs are small when
compared to solver costs (low cost of adding
nests)
• Model code must be further optimized and
physics calls must be reduced to attain further
speed-up beyond the saturation point
– Investigating ways to further reduce cost of 2-
way interaction
Slide courtesy: Thiago Quirino, HRD/AOML
Slide 11 of 29
Project Statement: “Test and evaluate the Basin-scale HWRF’s multi-storm initialization capabilities
in the NMMB/NEMS framework and assess potential for demonstrating the initialization real-time in
FY16 and implementing it in FY17.”
GFS cold-start
2015-09-03 00Z, MSLP Before Initialization
2015-09-03 00Z, MSLP After Initialization
After
initialization and
relocation
GFS cold-start
After
initialization and
relocation
IGNACIO12E JIMENA13E
SVN Repository path: https://svnemc.ncep.noaa.gov/projects/hnmmb/trunk/sorc/util/vortex_init_b
Multi-Storm Vortex Initialization and
Cycling in NMMB (EMC-HRD Collaborations supported by HIWPP and R2O/NGGPS)
Slide 12 of 29
Evaluation of the Hurricane NMMB (EMC-HRD Collaborations supported by HIWPP and R2O/NGGPS)
• Perform quasi real-time forecasts of Basin-scale NMMB at
18:06:02km resolution for multiple-storms in 2015
– Developed an end-to-end automation system for real-time forecasts
http://storm.aoml.noaa.gov/hnmmb
Slide courtesy: Thiago Quirino, HRD/AOML
Slide 13 of 29
1. The latest HWRF (2015) physics package has been added to
HNMMB, and tested.
2. HWRF moving nest algorithm has been implemented in
HNMMB, and tested
3. HNMMB Restart capability has been implemented.
4. HNNMB can be complied/run using PC-linux-based gfortran.
5. (1) , (2), (3), and (4) have been added to NMMB trunk
6. Post and tracker scripts are working with NMMB, and tested.
7. Rocoto workflow is being built. Main part of it (NPS,
relocation, NMMB, post, tracker) is working now. Lin is still
working on pulling data, and archiving..
8. Start to change NMMB scripts from ksh to python. Keqin is
leading this effort.
Progress of HNMMB at EMC
Slide 14 of 29
Red: HNMMB, no ocean, no DA
Purple: HWRF, no ocean, no DA
Amazing intensity/Pmin forecast from
HNMMB.
Performance of HNMMB for Hurricane
Patricia (2015)
Slide 15 of 29
Old algorithm: difficult to follow storms
when vortex is not well organized or
near islands.
New algorithm: follow the storm
HNNMB Nest motion algorithm
Slide 16 of 29
Options 2014 2015
microphys Fer Fer_hires
shortwave GFDL RRTM
longwave GFDL RRTM
turbulence GFSHUR GFSHUR (updated)
convection SASHUR SASHUR (scale-aware)
Sfc_layer GFDL GFDL
Land_surface GFDL noah
Physics options in HNMMB
Red: a new option or updated
Slide 17 of 29
Vertical X-Z cross section of wind: With 2015 hwrf phys, size and intensity
forecasts improved.
Impact of Physics on Hurricane Structure
Slide 18 of 29
Two-Way Nesting Capabilities in GFDL FV3 (Recent developments using HiRAM and FV3)
Year-long nonhydrostatic HiRAM
simulation using 2005 SSTs, using an 8-
km nest over the tropical Atlantic
Examples of high-resolution
nested grid simulations using
HiRAM and FV3
three-day HiRAM
forecasts of severe
convection during the
Moore, OK tornado
outbreak of May 2013,
in a simulation nesting
down to 1.3 km over
the southern plains
(using HIWPP 3km
global runs)
Slide courtesy: Lucas Harris, GFDL
Slide 19 of 29
Two-Way Nesting Capabilities in GFDL FV3 (Recent developments using HiRAM and FV3)
• Two-way nests in FV3 designed for simultaneous, consistent, coupled
regional and global solutions
Progress since last July:
• Improvements to nested grid algorithm to improve stability and accuracy of
boundary conditions, especially in regions of steep terrain.
• Simple renormalization-based update to ensure conservation of tracer
mass, which can now be conservatively updated to the coarse grid
• Maintenance of bitwise-reproducibility of solutions when changing
processor counts or restarting
• Started experimental sub-seasonal to seasonal (S2S) predictions in HiRAM
using a 8-km nested grid over the tropical Atlantic, based on existing Chen
and Lin seasonal prediction system
Slide courtesy: Lucas Harris, GFDL
Slide 20 of 29
North
American
refinement
MPAS: Mesh Generation: Lloyd’s Method
(iterative, using a user supplied density function)
Equatorial
refinement
Andes
refinement
Slide courtesy: Bill Skamarock, NCAR
Slide 21 of 29
MPAS: Global Mesh and Integration Options
Global Uniform Mesh Global Variable Resolution Mesh Regional Mesh - driven by
(1) previous global MPAS run (no spatial interpolation needed!)
(2) other global model run
(3) analyses
Voronoi meshes allows us to cleanly incorporate both
downscaling and upscaling effects (avoiding the problems in
traditional grid nesting) & to assess the accuracy of the
traditional downscaling approaches used in regional climate
and NWP applications.
Slide courtesy: Bill Skamarock, NCAR
Slide 22 of 29
MPAS-Atmosphere 2013-2014-2015
Tropical Cyclone Forecast Experiments
daily 10-day forecasts during the NH tropical cyclone season
Forecast Experiments with Variable-
Resolution Meshes
Western Pacific basin mesh Eastern Pacific basin mesh Atlantic basin mesh
Slide courtesy: Bill Skamarock, NCAR
Slide 23 of 29
Mesh Examples for NEPTUNE/NUMA
NEPTUNE: Navy Environmental Prediction sysTem Utilizing the NUMA2 corE
NUMA: Nonhydrostatic Unified Model of the Atmosphere (F. Giraldo NPS)
NEPTUNE-NUMA has a very flexible core that allows for static mesh
refinement, cubed-sphere, icosahedral meshes, limited area meshes.
Slide courtesy: Jim Doyle, NRL
Slide 24 of 29
NEPTUNE-NUMA Adaptive Mesh Refinement
•Non-conforming adaptive mesh
refinement (AMR) capability in
NEPTUNE will increase efficiency
•Possible applications: tropical
cyclones, dispersion, urban,
coastal, severe storms…
E. Hendricks
2-D Rising Bubble 2-D Vortex
Kopera and Giraldo JCP (2013) Slide courtesy: Jim Doyle, NRL
Slide 25 of 29
Variable Resolution Capabilities for NGGPS:
Phase 2 Testing
• Purpose is to demonstrate a baseline capability to provide
enhanced resolution over certain regions of interest, especially
for hurricanes and convective systems
• Approximately a 4:1 variation in horizontal resolution (3 km in
the vicinity of convective systems including hurricanes, up to 13
km in the far field) Individual groups can configure as they
choose, using fixed or moving high-resolution region, 1-way or
2-way nests
• Groups will be required to run the test with GFS physics, but
may submit supplementary tests with their own physics (since
‘scale-aware’ physics may be desirable in this case)
Slide 22 of 31
Slide 26 of 29
NGGPS Nesting Team Milestones and
Deliverables (2015-2016)
• NCEP/AOML: – Transition HIWPP funded upgrades, including ports of the HWRF nest movement
algorithm and idealized tropical cyclone framework into NEMS repository
– Determine the efficiency and scalability of NMMB/NEMS with two-way interactive nests
– Develop, test, and evaluate generalized grid-independent interpolation techniques for
free-standing nests in the NMMB/NEMS framework
– Test and evaluate the basin-scale HWRF multi-storm initialization in NMMB/NEMS
framework and assess potential for demonstrating the initialization real-time in FY16
and implementing it in FY17
– Implement scale-aware physics in NMMB for multi-storm multiple nest applications
– Develop preliminary capabilities for atmosphere-land-ocean-wave coupled system for
hurricane applications
– Proof of concept of global to local scale modeling system for hurricane predictions
• GFDL: – Subseasonal hurricane prediction in a prototype variable-resolution global NGGPS
model
• PSU: – Advancing Storm-Scale Forecasts over Nested Domains for High-Impact Weather
Green: Completed; Red: Ongoing
Slide 27 of 29
• Selected two best cases: 28 April 2014 (153 tornado
reports from Mississippi to Kentucky) and 6 May 2015
(65 tornado reports from Texas to Nebraska) based on
reasonable performance of NAM 4-km nests (location
and timing of CI and subsequent evolution of storms)
• Rerun fire weather nests at 1.33 km resolution (support
from SPC forecasters and Geoff DiMego/Eric Rogers,
EMC)
• Transferred datasets (>2TB) to PSU from EMC
• Developed diagnostic tools to evaluate the case studies
NGGPS Nesting Team: Updates from PSU Advancing Storm-Scale Forecasts over Nested Domains for
High Impact Weather
Slide 28 of 29
• Continue to increase resolution of nests that can operate
at cloud-resolving scales
• Couple nesting capability with more components as
added to NEMS
• Demonstrate global models operating at cloud resolving
scales with high-resolution nests for more accurate
forecasts of significant weather events
• Develop advanced post-processing techniques,
products, verification and diagnostic tools.
• Close interactions with other NGGPS atmospheric
dynamics, physics, data assimilation, overarching
system, software architecture and engineering teams
NGGPS Nesting Team Long-Term
Objectives
Slide 29 of 29
Questions?